Classification of Seizure Through SVM Based Classifier

Epilepsy is the most prevalent neural disorder characterized by abrupt and repetitive impairment of brain known as seizure, whose clinical symptoms are hyper synchronous activities of nerve cells in the brain. Since seizure, in general, occur very infrequently it is highly recommended for self-regulated disclosure of it during longstanding EEG measurement. The data handled in our work is publicly accessible online comprising of five classes. The segments in data set for each class were partitioned into two parts. Former comprised first 16 seconds (about 68%) of EEG which was used to train the network and rest of the signal were marked as test data. Statistical features for each class were evaluated. A supervised learning algorithm which is mostly used for classification and regression known as Support Vector Machine was used for classification of each set representing the different class. The classification results achieved taking all the five class was 91.42%. In order to confirm the accuracy of the classifier, the classifier is tested for different classification problem that were reported previously.